Statistical Regularities in ATM: Network Properties ... · 100 101 102 0 0.5 1 EGKK Degree Btwn,...

Post on 21-Jun-2020

1 views 0 download

Transcript of Statistical Regularities in ATM: Network Properties ... · 100 101 102 0 0.5 1 EGKK Degree Btwn,...

Statistical Regularities in ATM: NetworkProperties, Trajectory Deviations, and Delays

SID 2012 – Braunschweig, 27th of November, 2012

S. Vitali, G. Gurtner, L. Valori , M. Cipolla, V. Beato, S.Pozzi, S. Micciche, F. Lillo & R. Mantegna

ELSAEmpirically grounded agent based models for the future ATM scenario

Presentation of ELSA

Empirically grounded agent based models for the future ATM scenario

Deep BlueValentina Beato

Simone Pozzi

Universita diPalermo

Stefania VitaliMarco Cipolla

Salvatore MiccicheRosario Mantegna

Scuola NormaleSuperioreLuca Valori

Gerald GurtnerFabrizio Lillo

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 2 / 30

Presentation of ELSA

Aim

Build an Agent-Based Model integrating many actors at different levels inorder to test new scenarios in ATM.

Steps

Extract statistical regularities and stylized facts from traffic data,

build the ABM,

use regularities for calibrating and validating the future ABM.

Here we present a selection of empirical results extracted from the data.

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 3 / 30

Presentation of ELSA

Aim

Build an Agent-Based Model integrating many actors at different levels inorder to test new scenarios in ATM.

Steps

Extract statistical regularities and stylized facts from traffic data,

build the ABM,

use regularities for calibrating and validating the future ABM.

Here we present a selection of empirical results extracted from the data.

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 3 / 30

Data and Database

Data...

... on trajectories: M1 (last filled flight plan) and M3 (trajectoriesupdated by radar track) files containing sequences of navpoints,

... on the structure of Airspace (NEVAC files): sectors, routes, etc,

... for 16 AIRAC cycles ' 1 year and three months.

Database...

... eliminating redundancies,

... allowing very fast query on huge amount of data,

... building data of higher level (measure of complexity...).

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 4 / 30

Outline

1 AirportsNetworkStrength/degree distributionsNetwork communitiesDynamics

2 SectorsNetworkDynamicsDeviations

3 Navigation pointsCommunitiesDeviations

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 5 / 30

Airports

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 6 / 30

Airports: Network

Properties of nodes

Degree: number of destination from/to the airport

Strength: number of flights from/to the airport

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 7 / 30

Airports: Network

Properties of nodes

Degree: number of destination from/to the airport

Strength: number of flights from/to the airport

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 7 / 30

Airports: Network

Size proportional to degree.

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 8 / 30

Airports: network metrics

101 102 103100

101

102

PDF

= 2.2

Airport Strength100 101 102100

101

102

PDF

= 1.9

Airport Degree

Scale free network

Presence of hubs,

very short path between any points in the network (' 3).

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 9 / 30

Airports: companies

100 101 1020

0.5

1EDDM

Degree Bt

wn,

DLH

100 101 1020

0.5

1EGSS

Degree

Btw

n, R

YR

100 101 1020

0.5

1EKCH

Degree

Btw

n, S

AS

100 101 1020

0.5

1LEPA

Degree

Btw

n, B

ER

100 101 1020

0.5

1LTBA

Degree

Btw

n, T

HY

100 101 1020

0.5

1EGLL

Degree

Btw

n, B

AW

100 101 1020

0.5

1EGKK

Degree

Btw

n, E

ZY

100 101 1020

0.5

1LIRF

Degree

Btw

n, A

ZA

100 101 1020

0.5

1LFPG

Degree

Btw

n, A

FRBetweenness centrality

Measure of how much the node is central in the network ' number of shortestpath passing through it.

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 10 / 30

Airports: communities

General purpose/methods

Communities are group of nodes which are mutually more highlyinterconnected than they are with the rest of the network.

Many algorithms have been proposed to partition a network incommunities:

1 Maximization of modularity: finds the partition that maximizesmodularity, which is the fraction of the links within the givencommunities minus the expected such fraction if links were distributedat random (under some null hypothesis);

2 Infomap: based on random walks on networks,

After a partition has been obtained, one can characterize eachcommunity by measuring the over-expression of a given nodeattribute.

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 11 / 30

Airports: communities

Big supra national communities

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 12 / 30

Airports: communities

Big supra national communities

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 12 / 30

Airports: communities

Big supra national communities ⇒ tool to design airspaces?

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 12 / 30

Airports: dynamics

Seasonality

Number of flights and active airports are changing on a daily basis, yearlybasis and because of external shocks (volcano).

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 13 / 30

Airports: dynamics

Monday 13/05/2010 Sunday 19/05/2010

Cluster overexpressed size Cluster overexpressed sizeattribute attribute

1 uk 121 1 norway 1071 ireland 121 1 sweden 1071 france 121 1 finland 1072 norway 113 1 denmark 1072 sweden 113 2 uk 1042 finland 113 2 ireland 1042 denmark 113 3 germany-civil 973 germany-civil 89 3 austria 973 poland 89 3 romania 973 austria 89 4 spain 514 greece 45 4 portugal 514 romania 39 5 turkey 405 spain 39 6 italy 365 portugal 39 7 greece 326 turkey 347 italy 27

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 14 / 30

Sectors

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 15 / 30

Sectors: structure

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 16 / 30

Sectors: structure

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 17 / 30

Sectors: network

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 18 / 30

Sectors: strength and degree

Non scale-free network

typical degree, strength, length...

big diameter.

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 19 / 30

Sectors: dynamics

Structure of the network is changing during the day.

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 20 / 30

Sectors: dynamics

Change of the traffic on the network,

change of the underlying network (geographical neighbours ofsectors).

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 21 / 30

Sectors: deviations

The change in the structure allows to absorb the traffic without morereroutings.

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 22 / 30

Navigation points

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 23 / 30

Navpoints: map

Finer scale, geographical network.

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 24 / 30

Navpoints: communities

Infomap

Big communities, looking like airspaces: ACC?

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 25 / 30

Navpoints: communities

Infomap

Big communities, looking like airspaces: ACC?

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 25 / 30

Navpoints: communities

Infomap

Big communities, looking like airspaces: ACC? ⇒ tool to design airspaces?

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 25 / 30

Navpoints: deviations 1

Local metrics

Number of flights deviated,

point common in M1 (planned trajectory) and M3 (actual trajectory),

area generated,

etc...

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 26 / 30

Navpoints: deviations 2

Number of horizontal deviations drops with local traffic.

Number of vertical deviations increases with local traffic.

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 27 / 30

Navpoints: deviations 3

Nodes of low degree are more avoided when the traffic increases

Nodes of high degree are more avoided when the traffic decreases

Nodes of low degree gets flights more delayed when the trafficdecreases

Nodes of high degree gets flights more delayed when the trafficincreases

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 28 / 30

Conclusion

Airport Network

Scale-free network (small world),

different type of network for companies,

organized in communities which look like the FABs.

Sector Network

Geographical network,

dynamical structure on top of dynamical conditions (traffic).

Navpoint Network

Finer scale, geographical network,

organized in communities which look like the ACCs,

Deviations are handled differently at high degree nodes and lowdegree nodes.

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 29 / 30

Thanks for your attention

Deep BlueValentina Beato

Simone Pozzi

Universita diPalermo

Stefania VitaliMarco Cipolla

Salvatore MiccicheRosario Mantegna

Scuola NormaleSuperioreLuca Valori

Gerald GurtnerFabrizio Lillo

Gerald Gurtner et al. (ELSA) Statistical Regularities in ATM Braunschweig, Nov. 2012 30 / 30